Friday, June 23, 2006

AMD builds up Task Force Wipe Out

It's official. After announcing astonishing plans to expand capacity by 400% through FAB36 and FAB38 expansion in Dresden Germany, AMD now signed a deal to build a new FAB in New York.

This would allow AMD to completely displace IA32 with AMD64. Pax AMD is nearer. Prepare to hail Hector Caeser.

After K8L upgrade next year, AMD64 will be soon upgraded to Direct Connect Architecture 2.0 in 2008. Details on DCA 2.0 are lacking, but Hector Ruiz, CEO of Advanced Micro Devices, reckoned that it will be a real killer.

In other news, Henri Richard, EVP of AMD, pointed out the genetic defects of Intel, which I analysed previously here, here and here. Intel execs are definitely not bright. Cutting price by 60% to gain 3% market share is not just irrational, it's suicidal. I projected Intel's demise in 7 quarters.

The INQ reported about inverse threading we have been discussing.

13 Comments:

Anonymous Anonymous said...

Some guys in Bangalore do have some balls ::

AMD Exclusive Outlet :

http://www.channeltimes.com/channeltimes/jsp/article.jsp?article_id=74108&cat_id=741

I am surprised why folks out here dont see a cash cow in somethign like this !!
or will their family be wiped out the next day of opening ..lol

10:54 AM, June 23, 2006  
Anonymous Anonymous said...

"Direct Connect Architecture 2.0"

What is that? Something even better than HT3??

10:55 AM, June 23, 2006  
Anonymous Anonymous said...

How long does it take to build a FAB?

Looking at the new about Intels new FAB, it was just an addition to an existing FAB and it took them 2 years.

Thanks.

11:00 AM, June 23, 2006  
Blogger Sharikou, Ph. D said...

According to AMD's slides, 2007 will be the year for HT 3.0. For 2008, it's DCA 2.0.

11:02 AM, June 23, 2006  
Anonymous Anonymous said...

Can AMD afford a new plant and updating FAB30 at the same time as having a price war with Intel?

11:09 AM, June 23, 2006  
Blogger Sharikou, Ph. D said...

How long does it take to build a FAB?

Looking at the new about Intels new FAB, it was just an addition to an existing FAB and it took them 2 years.


AMD started building FAB36 in 2003. It started volume production in 1Q06.

11:12 AM, June 23, 2006  
Anonymous Richard said...

Hey Sharikou, do you mind dropping me an email kwan8168@tpg.com.au

Just can't seem to get in contact with you :P

11:22 AM, June 23, 2006  
Anonymous Anonymous said...

I noticed your earlier capacity analysis.

http://sharikou.blogspot.com/2006/03/amd-poised-to-exit-2006-with-55-market.html

The funny thing is how you deliberately underassume Intel's capacity. You are calculating for the end of 2006, and already Intel has 3 65nm Fabs operational not 2.

http://www.tgdaily.com/2006/06/22/intel_opens_leixlip_fab/

D1D has been in 65nm since the middle of last year I believe, Fab 12 was in 65nm production since November 2005, and now Fab 24-2 has been contributing 65nm production for 3 months now as it ramps. Between 3 full 65nm Fabs I think it's safe to assume that concerns over lack of volume for Core 2 is overrated. Especially considering there is still 1 month before Conroe launches and 2 months before Merom.

What's more, Intel has a fourth 65nm Fab D1C that is currently under conversion. It should be contributing production before the end of this year, just in time to supply the extra chips for quad cores.

In light of these facts, your earlier analysis that Intel only has 2 65nm Fabs by the end of 2006 in flawed. Intel will have 4. As a matter of comparison, AMD's single 65nm Fab, Fab 36 isn't scheduled to have 65nm match 90nm production until the end of 2006, and it won't be fully ramped until mid 2007. These are AMD's own optimistic estimates. What's more Chartered won't even begin ramping 65nm production until mid 2007.

Now in light of that, what does Intel have for 2007. First, D1D is going to be converted to 45nm production before the end of 2007. Plus, Intel already has 2 other 45nm Fabs under construction, Fab 32 and 28. These are both brand new ones which means they are adding additional capacity, not transferring it from older processes.

Now the funny thing is, that all this information was available in the Intel Manufacturing Map that you linked up in your earlier blog. What's more, you claimed that Intel only had 2 90nm Fabs when they actually have 4, Fab 24, Fab 18, Fab 11X, and D1D (as it transistions to 65nm). Between, the 90nm production Fabs, and the 65nm production Fabs, it seems that you are deliberately under-reporting Intel's capabilities despite the fact that they are clear from the PDF that you yourself linked up. The question is why?

11:42 AM, June 23, 2006  
Anonymous Anonymous said...

With AMD's archictecture all things are possible... Wonder when intel shareholders will figure this out.

1:14 PM, June 23, 2006  
Blogger Sharikou, Ph. D said...

The funny thing is how you deliberately underassume Intel's capacity. You are calculating for the end of 2006, and already Intel has 3 65nm Fabs operational not 2.


No. I don't know exactly what kind of capacity Intel has, because I don't know their exact yield. I simply assumed that Intel has the capacity to produce 100% of the world's demand. The number of FABs don't matter for Intel. The more FABs Intel has, it will only show Intel's yields are even lower, because Intel is producing only 4x the chips of AMD. AMD had only one 200mm FAB and was taking 22% of the market.

12:30 PM, June 24, 2006  
Anonymous Anonymous said...

http://www.theinquirer.net/?article=32617

ooops...

1:32 AM, June 25, 2006  
Anonymous Anonymous said...

you have any more explanation?

http://www.pcper.com/article.php?aid=265

1:33 AM, June 25, 2006  
Anonymous Anonymous said...

http://www.princeton.edu/~rblee/ELE
572Papers/DynamicMultithreadingProc
_akkary.pdf

looks like intel has silently worked on this before AMD hyped the public. it is rumored that such technology was implemented on Core2 processors. the technology, intel multiplexing technology, may have helped Conroe to achieve its performance. however, if not, then we will only see significant increase in Conroe's performance.

1:40 AM, June 25, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home